Essential Algorithms Research In Vehicle Detection

Now ITS (Intelligent Transport System) is an important research direction of computer vision. Comparison with the traditional methods, ITS possesses lots of advantages such as low cost, good performance and rich functions. The detection of video vehicle is the core of ITS which is the foundation of other ITS technologies. The technology of moving vehicle detection needs to be improved because it remains immature resulting from its complexity. In this thesis, we do some researches on the key problems of the moving vehicle detection, give new methods and demonstrate the validity by experiments. The main contents of the study include such aspects as follows:1. Since the clear input monitor sequence or image is the basic of the vehicle detection, we do the research of image enhancement in the Chapter Two. The thesis introduces two common algorithms (Histogram Equalization and Unsharp Enhancement) in the image enhancement. And we propose the algorithm using adaptive threshold and nonlinear plus arithmetic operator to solve the problem in enhancing image blocks. The proposed method gets better result in comparison with the two common ones.2. It is incapable for traditional methods to accurately detect the moving vehicle in the dynamic background because the interferential moving objects in dynamic scenes (e.g. changeable weather and the lighting conditions) have great influences on the accuracy of moving vehicle detection. By investigating the distinction between the motion vectors of the dynamic background and those of the moving vehicles, motion models are in possession of different distribution characters on the base of GMM. The proposed method set up motion models separately. Then, the intra-frame information is used to modify the motion model. Finally, Bayesian decision rule is implemented to detect vehicle from dynamic background. The experiment results show that the proposed method can detect moving vehicle more accurately and robustly in complex and dynamic scenes in comparison with other vehicle detection methods.3. There is a lot of information of vehicle in the intra-frame. To utilize the intra-frame information, especially to obtain the feature, is good complementarities to the vehicle detection based on the inter-frame information. And it can effectively improve the detail information of the proposed method above.